Btrfs: Create a work queue for bio writes
authorChris Mason <chris.mason@oracle.com>
Wed, 16 Apr 2008 15:14:51 +0000 (11:14 -0400)
committerChris Mason <chris.mason@oracle.com>
Thu, 25 Sep 2008 15:04:01 +0000 (11:04 -0400)
This allows checksumming to happen in parallel among many cpus, and
keeps us from bogging down pdflush with the checksumming code.

Signed-off-by: Chris Mason <chris.mason@oracle.com>
fs/btrfs/ctree.h
fs/btrfs/disk-io.c
fs/btrfs/disk-io.h
fs/btrfs/extent_io.h
fs/btrfs/inode.c

index a22edcf..ff15b85 100644 (file)
@@ -514,8 +514,11 @@ struct btrfs_fs_info {
        struct list_head hashers;
        struct list_head dead_roots;
        struct list_head end_io_work_list;
+       struct list_head async_submit_work_list;
        struct work_struct end_io_work;
+       struct work_struct async_submit_work;
        spinlock_t end_io_work_lock;
+       spinlock_t async_submit_work_lock;
 
 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
        struct work_struct trans_work;
index 79c284c..9e41ea9 100644 (file)
@@ -47,6 +47,7 @@ static int check_tree_block(struct btrfs_root *root, struct extent_buffer *buf)
 
 static struct extent_io_ops btree_extent_io_ops;
 static struct workqueue_struct *end_io_workqueue;
+static struct workqueue_struct *async_submit_workqueue;
 
 struct end_io_wq {
        struct bio *bio;
@@ -58,6 +59,15 @@ struct end_io_wq {
        struct list_head list;
 };
 
+struct async_submit_bio {
+       struct inode *inode;
+       struct bio *bio;
+       struct list_head list;
+       extent_submit_bio_hook_t *submit_bio_hook;
+       int rw;
+       int mirror_num;
+};
+
 struct extent_map *btree_get_extent(struct inode *inode, struct page *page,
                                    size_t page_offset, u64 start, u64 len,
                                    int create)
@@ -365,7 +375,31 @@ int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio,
        return 0;
 }
 
-static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
+int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode,
+                       int rw, struct bio *bio, int mirror_num,
+                       extent_submit_bio_hook_t *submit_bio_hook)
+{
+       struct async_submit_bio *async;
+
+       async = kmalloc(sizeof(*async), GFP_NOFS);
+       if (!async)
+               return -ENOMEM;
+
+       async->inode = inode;
+       async->rw = rw;
+       async->bio = bio;
+       async->mirror_num = mirror_num;
+       async->submit_bio_hook = submit_bio_hook;
+
+       spin_lock(&fs_info->async_submit_work_lock);
+       list_add_tail(&async->list, &fs_info->async_submit_work_list);
+       spin_unlock(&fs_info->async_submit_work_lock);
+
+       queue_work(async_submit_workqueue, &fs_info->async_submit_work);
+       return 0;
+}
+
+static int __btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
                                 int mirror_num)
 {
        struct btrfs_root *root = BTRFS_I(inode)->root;
@@ -389,6 +423,17 @@ static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
        return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio, mirror_num);
 }
 
+static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
+                                int mirror_num)
+{
+       if (!(rw & (1 << BIO_RW))) {
+               return __btree_submit_bio_hook(inode, rw, bio, mirror_num);
+       }
+       return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info,
+                                  inode, rw, bio, mirror_num,
+                                  __btree_submit_bio_hook);
+}
+
 static int btree_writepage(struct page *page, struct writeback_control *wbc)
 {
        struct extent_io_tree *tree;
@@ -903,9 +948,9 @@ static int bio_ready_for_csum(struct bio *bio)
 }
 
 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
-void btrfs_end_io_csum(void *p)
+static void btrfs_end_io_csum(void *p)
 #else
-void btrfs_end_io_csum(struct work_struct *work)
+static void btrfs_end_io_csum(struct work_struct *work)
 #endif
 {
 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
@@ -959,6 +1004,39 @@ void btrfs_end_io_csum(struct work_struct *work)
        }
 }
 
+#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
+static void btrfs_async_submit_work(void *p)
+#else
+static void btrfs_async_submit_work(struct work_struct *work)
+#endif
+{
+#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
+       struct btrfs_fs_info *fs_info = p;
+#else
+       struct btrfs_fs_info *fs_info = container_of(work,
+                                                    struct btrfs_fs_info,
+                                                    async_submit_work);
+#endif
+       struct async_submit_bio *async;
+       struct list_head *next;
+
+       while(1) {
+               spin_lock(&fs_info->async_submit_work_lock);
+               if (list_empty(&fs_info->async_submit_work_list)) {
+                       spin_unlock(&fs_info->async_submit_work_lock);
+                       return;
+               }
+               next = fs_info->async_submit_work_list.next;
+               list_del(next);
+               spin_unlock(&fs_info->async_submit_work_lock);
+
+               async = list_entry(next, struct async_submit_bio, list);
+               async->submit_bio_hook(async->inode, async->rw, async->bio,
+                                      async->mirror_num);
+               kfree(async);
+       }
+}
+
 struct btrfs_root *open_ctree(struct super_block *sb,
                              struct btrfs_fs_devices *fs_devices)
 {
@@ -987,14 +1065,17 @@ struct btrfs_root *open_ctree(struct super_block *sb,
        }
        end_io_workqueue = create_workqueue("btrfs-end-io");
        BUG_ON(!end_io_workqueue);
+       async_submit_workqueue = create_workqueue("btrfs-async-submit");
 
        INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_NOFS);
        INIT_LIST_HEAD(&fs_info->trans_list);
        INIT_LIST_HEAD(&fs_info->dead_roots);
        INIT_LIST_HEAD(&fs_info->hashers);
        INIT_LIST_HEAD(&fs_info->end_io_work_list);
+       INIT_LIST_HEAD(&fs_info->async_submit_work_list);
        spin_lock_init(&fs_info->hash_lock);
        spin_lock_init(&fs_info->end_io_work_lock);
+       spin_lock_init(&fs_info->async_submit_work_lock);
        spin_lock_init(&fs_info->delalloc_lock);
        spin_lock_init(&fs_info->new_trans_lock);
 
@@ -1041,9 +1122,12 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 
 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
        INIT_WORK(&fs_info->end_io_work, btrfs_end_io_csum, fs_info);
+       INIT_WORK(&fs_info->async_submit_work, btrfs_async_submit_work,
+                 fs_info);
        INIT_WORK(&fs_info->trans_work, btrfs_transaction_cleaner, fs_info);
 #else
        INIT_WORK(&fs_info->end_io_work, btrfs_end_io_csum);
+       INIT_WORK(&fs_info->async_submit_work, btrfs_async_submit_work);
        INIT_DELAYED_WORK(&fs_info->trans_work, btrfs_transaction_cleaner);
 #endif
        BTRFS_I(fs_info->btree_inode)->root = tree_root;
@@ -1403,6 +1487,9 @@ int close_ctree(struct btrfs_root *root)
        flush_workqueue(end_io_workqueue);
        destroy_workqueue(end_io_workqueue);
 
+       flush_workqueue(async_submit_workqueue);
+       destroy_workqueue(async_submit_workqueue);
+
        iput(fs_info->btree_inode);
 #if 0
        while(!list_empty(&fs_info->hashers)) {
index 60b0190..5d5f697 100644 (file)
@@ -74,4 +74,7 @@ int btrfs_verify_block_csum(struct btrfs_root *root,
                            struct extent_buffer *buf);
 int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio,
                        int metadata);
+int btrfs_wq_submit_bio(struct btrfs_fs_info *fs_info, struct inode *inode,
+                       int rw, struct bio *bio, int mirror_num,
+                       extent_submit_bio_hook_t *submit_bio_hook);
 #endif
index b47859c..232300d 100644 (file)
 
 struct extent_state;
 
+typedef        int (extent_submit_bio_hook_t)(struct inode *inode, int rw,
+                                      struct bio *bio, int mirror_num);
 struct extent_io_ops {
        int (*fill_delalloc)(struct inode *inode, u64 start, u64 end);
        int (*writepage_io_hook)(struct page *page, u64 start, u64 end);
-       int (*submit_bio_hook)(struct inode *inode, int rw, struct bio *bio,
-                              int mirror_num);
+       extent_submit_bio_hook_t *submit_bio_hook;
        int (*merge_bio_hook)(struct page *page, unsigned long offset,
                              size_t size, struct bio *bio);
        int (*readpage_io_hook)(struct page *page, u64 start, u64 end);
index 7869c01..457daf7 100644 (file)
@@ -324,13 +324,29 @@ int btrfs_merge_bio_hook(struct page *page, unsigned long offset,
        return 0;
 }
 
-int btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
+int __btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
                          int mirror_num)
 {
        struct btrfs_root *root = BTRFS_I(inode)->root;
        struct btrfs_trans_handle *trans;
        int ret = 0;
 
+       mutex_lock(&root->fs_info->fs_mutex);
+       trans = btrfs_start_transaction(root, 1);
+       btrfs_set_trans_block_group(trans, inode);
+       btrfs_csum_file_blocks(trans, root, inode, bio);
+       ret = btrfs_end_transaction(trans, root);
+       BUG_ON(ret);
+       mutex_unlock(&root->fs_info->fs_mutex);
+       return btrfs_map_bio(root, rw, bio, mirror_num);
+}
+
+int btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
+                         int mirror_num)
+{
+       struct btrfs_root *root = BTRFS_I(inode)->root;
+       int ret = 0;
+
        if (!(rw & (1 << BIO_RW))) {
                ret = btrfs_bio_wq_end_io(root->fs_info, bio, 0);
                BUG_ON(ret);
@@ -342,13 +358,9 @@ int btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
                goto mapit;
        }
 
-       mutex_lock(&root->fs_info->fs_mutex);
-       trans = btrfs_start_transaction(root, 1);
-       btrfs_set_trans_block_group(trans, inode);
-       btrfs_csum_file_blocks(trans, root, inode, bio);
-       ret = btrfs_end_transaction(trans, root);
-       BUG_ON(ret);
-       mutex_unlock(&root->fs_info->fs_mutex);
+       return btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info,
+                                  inode, rw, bio, mirror_num,
+                                  __btrfs_submit_bio_hook);
 mapit:
        return btrfs_map_bio(root, rw, bio, mirror_num);
 }